I have access to high frequency data for a few instruments using which I can simulate a limit order book.I would like to predict direction of price(best bid/ask) in the short term(1 sec, 5 sec and 10 sec) using that. What would be a good model/reference point to start with? Additionally, if there are any research papers/books on similar problems, please let me know.
$\begingroup$
$\endgroup$
Add a comment
|
2 Answers
$\begingroup$
$\endgroup$
1
If you have access to full order log from a trading venue, you can build (not simulate) the actual limit order book, with tick by tick changes. Basic top-of-book mispricing is arbitraged away within 500 μs which includes getting the market data, updating the book, performing analysis, issuing an order and delivering it to the exchange. Check out IEX SEC filings and the rule book for CQI formula to get a sense of moving parts.
-
$\begingroup$ Apologies for using the term simulate loosely. I meant I can recreate the order book with my dataset but wanted to know a starting point for my research $\endgroup$ Commented May 13, 2021 at 12:46
$\begingroup$
$\endgroup$
Your question is quite broad, but I try my best to give you some hints to tackle this:
To predict the price direction, you need to build a signal from your order book. I recommend Information-driven bars like Tick Imbalanced Bars or Volume Imbalanced Bars. Then you can run a LSTM or something to get a good prediction of the next incoming orders.
However, there are many many ways to construct signals from the order book, so depends on your intend. If you have access to the cancelled orders as well, you can calculate: market impact = |execution price - bendchmark price| * shares executed at execution price
You can also calculate some volume-weighted average price (VWAP) or the imlementation shortfalls to get some signals from the order book.
Good approaches can be found here:
- De Prado, M. L. (2018). Advances in financial machine learning. John Wiley & Sons.
- Jansen, S. (2020). Machine Learning for Algorithmic Trading: Predictive Models to Extract Signals from Market and Alternative Data for Systematic Trading Strategies with Python. Packt Publishing Limited.
